Dartmouth College Dartmouth Digital Commons

Open Dartmouth: Faculty Open Access Articles

4-2017 and Percent Saturation Levels Predict Type 2 Diabetes Risk and Cardiovascular Disease Outcomes Leo R. Zacharski Dartmouth College, [email protected]

Galina Shamayeva Veterans Affairs Palo Alto Health Care System

Bruce K. Chow Veterans Affairs Palo Alto Health Care System

Ralph G. DePalma Veterans Affairs Office of Research and Development, Washington DC

Follow this and additional works at: https://digitalcommons.dartmouth.edu/facoa Part of the Endocrinology, Diabetes, and Commons

Recommended Citation Zacharski, Leo R.; Shamayeva, Galina; Chow, Bruce K.; and DePalma, Ralph G., "Ferritin and Percent Transferrin Saturation Levels Predict Type 2 Diabetes Risk and Cardiovascular Disease Outcomes" (2017). Open Dartmouth: Faculty Open Access Articles. 433. https://digitalcommons.dartmouth.edu/facoa/433

This Article is brought to you for free and open access by Dartmouth Digital Commons. It has been accepted for inclusion in Open Dartmouth: Faculty Open Access Articles by an authorized administrator of Dartmouth Digital Commons. For more information, please contact [email protected]. Send Orders for Reprints to [email protected] Current Diabetes Reviews, 2017, 13, 00-00 1

RESEARCH ARTICLE Ferritin and Percent Transferrin Saturation Levels Predict Type 2 Diabetes Risk and Cardiovascular Disease Outcomes

Leo R. Zacharski1,*, Galina Shamayeva2, Bruce K. Chow2 and Ralph G. DePalma3

1Research Service, Department of Veterans Affairs Medical Center, White River Jct., VT and the Department of Medicine, Geisel School of Medicine at Dartmouth College, Lebanon, NH, USA; 2Veterans Affairs Cooperative Studies Program, Research Service, Veterans Affairs Palo Alto Health Care System, Palo Alto, CA, USA; 3Veterans Affairs Office of Research and Development, Washington DC and the Norman Rich Department of Surgery, Uniformed Services University of the Health Sciences, Bethesda, MD, USA

Abstract: Introduction: Type 2 diabetes (T2D) and cardiovascular disease (CVD) risk associate with ferritin and percent transferrin saturation (%TS) levels. However, increased risk has been observed at levels considered within the “normal range” for these markers. Objective: To define normative ferritin and %TS levels associated with T2D and CVD risk. Methods: Six-monthly ferritin, %TS and hemoglobin levels from 1,277 iron reduction clinical trial A R T I C L E H I S T O R Y participants with CVD (peripheral arterial disease, 37% diabetic) permitted pair-wise analysis using Loess Locally Weighted Smoothing plots. Curves showed continuous quantitative ferritin, hemoglobin Received: March 23, 2017 Revised: April 25, 2017 (reflecting physiologic iron requirements), and %TS (reflecting iron transport and sequestration) levels Accepted:April 27, 2017 over a wide range of values. Inflection points in the curves were compared to ferritin and %TS levels

DOI: indicating increased T2D and CVD risk in epidemiologic and intervention studies. 10.2174/1573399813666170504163138 Results: Increasing ferritin up to about 80 ng/mL and %TS up to about 25% TS corresponded to in- creasing hemoglobin levels, and minimal T2D and CVD risk. Displaced Loess trajectories reflected lower hemoglobin levels in diabetics compared to non-diabetics. Ferritin levels up to about 100 ng/mL paralleled proportionately increasing %TS levels up to about 55%TS corresponding to further limita- tion of T2D and CVD risk. Ferritin levels over 100 ng/mL did not associate with hemoglobin levels and coincided with increased T2D and CVD risk. Conclusions: Recognition of modified normal ranges for ferritin from about 15 ng/mL up to about 80- 100 ng/mL and %TS from about 15% up to about 25-55% may improve the value of iron biomarkers to assess and possibly lower T2D and CVD risk. Keywords: Iron, ferritin, transferrin, diabetes, normal values.

INTRODUCTION documented continuous linear dose-related associations be- tween ferritin levels and increased T2D risk in men and Diabetes is associated with bidirectional abnormalities of women with highly significant differences across ferritin glucose and iron metabolism [1-5]. Meta-analyses of obser- quintiles. Continuously increasing hazard ratios began at vational studies have shown that ferritin levels in the general ferritin levels above about 100 ng/mL, a level generally con- population predict type 2 diabetes (T2D) and gestational sidered to be within the normal range. Quintile boundaries diabetes mellitus (GDM) risk [6-11]. A longitudinal epide- placed rising risk in the second quintile for males having low miologic study showed dose-over-time relationships between and high cut points of 68 ng/mL and 117 ng/mL respec- elevated ferritin levels and increased mortality from diabetes tively, and in the fourth risk quintile for women having cut and other age-related diseases [12]. Podmore and colleagues points of 73 ng/mL and 121 ng/mL respectively. In contrast, reported the predictive value of measures of iron status reduced T2D risk with higher percent transferrin saturation for subsequent diabetes risk [13]. These investigators (%TS) levels was observed in women at a cutoff approxi- mating 45%TS. This association diminished above 50%TS *Address correspondence to this author at the Research Service, Department and was lost at a cutoff above 55%TS. These findings were of Veterans Affairs Medical Center, White River Jct., Vermont, USA, consistent with progressively reduced capacity for transferrin Zip code 05009; Tel: 603-650-6763; Fax: 603-650-6786; binding of excess iron or redox-active non-transferrin bound E-mail: [email protected] iron (NTBI) at %TS levels above about 50-55% [13, 14]. A

1573-3998/17 $58.00+.00 © 2017 Bentham Science Publishers 2 Current Diabetes Reviews, 2017, Vol. 13, No. 0 Zacharski et al. similar trend in men did not reach statistical significance pitals and a national IRB approved the protocol. Consent- possibly due to overall higher ferritin levels in men [14-16]. ing (98.8% male) patients over age 21 (average age 67) with stable CVD (peripheral arterial disease) were com- We observed that higher ferritin levels correlated with puter randomized by diabetes and smoking status, age, fer- mortality during a prospective randomized clinical trial of iron reduction in primarily males with cardiovascular disease ritin level, HDL/LDL ratio, and medical center [28-30]. Thirty-seven percent of the 1,277 subjects had a (CVD, peripheral arterial disease, VACSP 410) [17] along clinical diagnosis of diabetes at entry. This study was de- with increased levels of inflammatory markers: TNF alpha signed to optimize opportunity to detect an effect of fol- (receptor 1), IL6 and hsCRP [18, 19]. Significantly improved low-up ferritin levels on the primary outcome, all-cause disease outcomes related to lower ferritin levels [17, 20, 21] mortality, and the secondary outcome, death plus non-fatal and with increasing %TS levels in the range of 25-50%TS [21]. Mean ferritin levels in participants in CSP 410 at entry myocardial infarction and stroke. Patients were required to have stable CVD for the past 6 months, no disorder of iron were about 122 ng/mL and iron reduction targeted ferritin metabolism including iron deficiency or hemochromatosis, levels of 60 ng/mL [17]. Ferritin levels in participants having no malignancy within the past 5 years, hematocrit of >35%, no mortality, or non-fatal myocardial infarction or stroke, ferritin level of <400 ng/mL, and creatinine <3.0 mg/dL over 6-years of follow-up averaged 76 ng/mL [17]. In a par- [28-30]. ferritin levels up to 400 ng/mL, encoun- ticipating institution sub-study, the mean ferritin level corre- sponding to mortality was 132.5 ng/mL while the mean fer- tered commonly in practice [12], were similar for diabetics and non-diabetics (see results section). Because diabetes ritin in those with no mortality was 82.6 ng/mL [18, 19]. was a randomization variable, equivalent numbers of dia- These steep dose-effect levels approximated ferritin levels betics and non-diabetics were randomized to iron reduction associated with reduced morbidity and increased longevity versus control groups. The use of calibrated phlebotomy observed in other population studies [20, 22-25], as well as based on ferritin levels (ml blood to be removed = (ferritin threshold levels of ferritin and protective %TS levels ob- served by for T2D risk by Podmore and associates [13]. level-25) X10) for iron reduction provided a wide range of ferritin, %TS and hemoglobin values observed in labora- These epidemiologic and clinical trial associations sug- tory test panels obtained at 6-monthly intervals during six gested revision of normative values for ferritin and %TS years of observation. Ferritin and %TS levels reported here levels, common measures of iron metabolism, considered represent the 6-monthly (maximum) levels measured dur- indicative of the transition from low to increased disease ing follow-up used to calculate the need for further phle- risk. Similarities in iron biomarker levels associated with botomy. For purposes of this report, quantitative data at T2D risk [13, 15] and adverse vascular disease outcomes entry were examined for normality using the Shapiro-Wilk [17, 21] suggested that examination of graded biomarker test [31] and found to be not normally distributed. Thus, levels during iron reduction would provide further informa- quantitative data are presented below as medians with 25th tion on quantitative relationships between these biomarkers and 75th percentiles. Statistical significance between of iron metabolism and disease risk. groups was tested using the Wilcoxon test [32]. This report demonstrates quantitative relationships be- Loess Locally Weighted Scatter Smoothing Plot analysis tween a wide range of levels of ferritin, %TS and hemoglo- [26] was used to detect interrelationships between these vari- bin (representing physiologic iron) obtained by direct clini- ables. This method permits construction of a smooth curve in cal measurement using Loess plots [26] to display a transi- order to visualize relationships between two variables over a tion from physiologic to pathologic markers of iron metabo- continuum not otherwise recognizable within large popula- lism predictive of adverse clinical outcomes. We reported tions [33-35]. Inflection point relationships between two previously the utility of Loess methodology to characterize variables of interest allow identification of subpopulations racial variation in red cell and iron homeostasis [27]. Our within a larger clinical trial population with unique charac- present findings clarify normative ferritin and %TS levels as teristics associated with graded risk of adverse clinical out- measures of iron metabolism predictive of lower disease comes [36, 37]. Measures of interest included relationships risk. between ferritin, %TS reflecting iron transport and seques- tration, and hemoglobin levels reflecting physiologic iron MATERIALS AND METHODS status. Loess methodology has been used to show relation- ships between individual laboratory variables and clinical The Iron and Atherosclerosis Study, FeAST, VA Coop- outcomes in studies of diabetes [36-39] and CVD [27, 40, erative Study Program # 410 (CSP 410); a prospective, 41]. The present study plotted relationships between ferritin randomized, controlled single-blinded clinical trial of iron and hemoglobin levels (8388 paired samples, Fig. 1), be- reduction; used graded phlebotomy based on ferritin levels tween %TS and hemoglobin levels (8371 paired samples, (Trial Registration: www.clinicaltrials.gov, Identifier Fig. 2) and between ferritin and %TS levels (8371 paired NCT00032357) [27-29] to test the hypothesis that im- samples, Fig. 3). Loess curves were examined for inflections proved clinical outcomes might be achieved in subjects below or above which adverse clinical events were less or with established CVD by reducing iron stores to levels more likely to occur. Relevance of observed inflection points typical of children and pre-menopausal women (about 25 to to clinical outcomes was based on outcomes from clinical 60 ng/mL). The Consort Diagram [28] and methodological trial findings [17, 21, 28] as well as epidemiologic studies details of study protocols; informed consent, randomization assessing iron biomarker levels and risk of T2D and CVD and intervention procedures; outcome assessment and study [12, 13, 16]. administration have been reported [28-30]. Institutional Review Boards (IRBs) at each of the 24 participating hos- Iron Measures, Diabetes and Cardiovascular Risk Current Diabetes Reviews, 2017, Vol. 13, No. 0 3

Fig. (3). Loess plots of 8371 paired ferritin and %TS values in dia- Fig. (1). Loess smoothing plots of 8388 paired values of hemoglo- betic and non-diabetic participants during the iron reduction trial, bin and ferritin in diabetic and non-diabetic participants during the CSP 410. The range of values is depicted in the axes of the figure. iron reduction trial, CSP 410. The range of values is depicted in the Rising curves overlap with an inflection point marked by the left- axes of the figure. Vertical arrows approximate inflection points hand vertical arrow at about 25%TS and about 80 ng/mL ferritin associated with the limits of hemoglobin iron requirement, and (similar to values shown in figures 1 and 2). The right-hand vertical altered T2D and CVD risk. arrow marks a second inflection point at about 55%TS and 100 ng/mL of ferritin.

RESULTS The FeAST study included 1,277 entrants, 473 of whom were diabetic and 804 non-diabetic at entry. Levels of fer- ritin, %TS and hemoglobin found at entry are expressed here as medians with 25th and 75th percentiles in parentheses. Me- dian levels of ferritin of 103 ng/mL (61 ng/mL, 170 ng/mL) and 103 (58, 168) ng/mL, (p=0.681); and median levels of %TS were 23.4% (18.7%, 30.1%) and 24.7% (18.8%, 30.8%), (p=0.141) were similar at entry for diabetics and non-diabetics respectively. Median entry levels of hemoglo- bin were slightly but statistically significantly lower in dia- betics compared to non-diabetics: 14.5 g/dL (13.5 g/dL, 15.3 g/dL) versus 14.8 g/dL (13.9 g/dL, 15.7 g/dL), (p<0.001). Loess plot trajectories for diabetic and non-diabetic partici- pants were similar but displaced for diabetics as shown in figures 1 and 2 owing to lower hemoglobin levels in diabet- ics. Fig. (1) shows Loess smoothing plots for hemoglobin and ferritin paired values. Levels of both increase in parallel up to an inflection point at ferritin levels approximating 80-100 ng/mL reflecting a possible mean upper limit for the physi- ologic “normal range” (represented by hemoglobin levels) Fig. (2). Loess plots of 8371 paired values of hemoglobin and %TS for both measures. Loess curves for diabetics and non- in diabetic and non-diabetic participants during the iron reduction diabetics were roughly parallel but displaced downward in trial, CSP 410. The range of values is depicted in the axes of the diabetics as predicted by the significantly lower hemoglobin figure. The vertical arrow shown here at about 20%TS approxi- levels in diabetic participants. The inflection point for the mates the inflection point (of 20-25%TS) representing transport hemoglobin level may be interpreted as occurring at a iron for erythropoiesis. somewhat lower ferritin level in diabetics compared to non- diabetics. However, ferritin levels below about 80-100 ng/mL followed a similar trajectory in diabetics and non-

4 Current Diabetes Reviews, 2017, Vol. 13, No. 0 Zacharski et al. diabetics. Ferritin levels above the transition at about 80-100 poration of iron in physiologic excess (or possibly redox- ng/mL lacked physiologic correspondence to increasing he- active NTBI) onto transferrin [14]. moglobin levels, an effect that was exaggerated in diabetics The proportionate increase in ferritin and %TS values and possibly contributory to “anemia” in certain diabetics shown in Figure 3 up to ferritin levels of about 80 ng/mL and [42]. %TS levels of about 25% suggests strict physiologic corre- Fig. (2) shows the Loess plots of paired hemoglobin and spondence or compensation of these functionally different %TS levels. As in Fig. (1), these plots also were roughly proteins representing balance between ferritin and transferrin parallel but displaced downward in diabetics. The inflection as they interact to maintain optimal (compensated) iron ho- point for hemoglobin levels depicted in Fig. (2) occurred at meostasis [14] associated with minimal disease risk [13, 17, about 20%TS in both diabetics and non-diabetics. Levels 21]. The curves flatten above ferritin levels of about 80 plateaued above about 25%TS and followed roughly similar ng/mL and %TS levels above about 25% but then increase trajectories in relation to hemoglobin levels in diabetics and proportionately up to about 100 ng/mL of ferritin of and non-diabetics throughout the range of values in this data set. about 50-55%TS. These relationships suggest sustained compensation by an increase in %TS as related to increasing Fig. (3) shows Loess plots of paired ferritin and %TS ferritin levels. Podmore and colleagues [13] also observed levels which increase synchronously with overlapping curves in diabetics and non-diabetics up to levels of about 55%TS. apparent clinical compensation for increasing ferritin levels with increasing %TS levels up to about 50-55%TS which The plateau above about 25%TS, the presumed cutoff for appeared to protect against T2D risk. The transition from physiologic levels of transport iron for erythropoiesis repre- physiologic to apparently de-compensated (disproportion- sented by the %TS level, is similar to that in Fig. (2). Curves ately increasing) levels above about 50-55%TS in women flatten but continue to overlap as levels of both rise in pro- was associated with increased T2D risk in that study [13]. portion to each other for ferritin levels up to about 100 ng/mL and %TS levels up to about 55%TS. However, curves We also noted a similar inflection point in our male partici- pants with CVD at about 25%TS (Figs. 2 and 3). The present for diabetics versus non-diabetics diverge at ferritin levels data suggest a transition from presumably protective to po- above about 100 ng/mL and %TS levels above about 55%. tentially toxic %TS levels above a range of about 25-55%TS Above these levels, dissociation of these measures appeared in males (Figs. 2 and 3). Progressive iron loading may ex- in which a further increase in ferritin levels is accompanied ceed the capacity for physiologic compensation by transfer- by a proportionate rise in %TS in non-diabetics while a less proportionate ferritin/%TS response occurred in diabetics. rin that may be most efficient when ferritin levels remain below about 80-100 ng/mL [17, 21]. Iron biomarker relation- ships shown in Figures 2 and 3 corresponding closely to the DISCUSSION epidemiologic findings of Podmore and colleagues [13] for Loess plots for diabetic and non-diabetic cohorts differed their observed hazard ratios for ferritin and %TS for T2D because of slightly lower hemoglobin values in diabetics risk. The apparent functional dissociation of ferritin and related presumably to “anemia of chronic disease” occurring %TS above these levels suggests a testable hypothesis that in some diabetics [42]. However, inflection points in relation transferrin may become occupied increasingly with poten- to hemoglobin levels occurred in both diabetic and non- tially toxic NTBI at ferritin levels considered previously to diabetic cohorts for ferritin levels at about 80-100 ng/mL, be normal [14, 43-45]. and for %TS levels at about 25% TS (Figs. 1 and 2, respec- Differing functions and regulatory control of ferritin as tively). Increasing ferritin levels up to these inflection points compared to %TS levels may also explain the positive suggest increasing physiologic incorporation of iron into association of ferritin levels with the inflammatory markers, hemoglobin. TNF alpha, IL6 and hsCRP, [18, 19, 43] as well as the Data presented in Fig. (1) suggest that ferritin levels ap- negative association of these inflammatory markers with proximating 80-100 ng/mL represent an upper normal limit %TS levels [42]. In accord with these findings, Podmore and of physiologic levels of body iron as related to correspond- colleagues also recorded that lower ferritin levels and higher ing hemoglobin levels. Ferritin levels over the 80-100 ng/mL %TS associate with lower hsCRP levels [13]. Side-by-side range appear to lack physiologic correspondence with hemo- comparison of quantitative %TS data reported here coincides globin levels. Thus, normative values for the serum ferritin with clinical outcome data reported in epidemiologic [13] may range from the commonly accepted lower limit of less and clinical trial outcome [21] data. Protective effects appear than about 15 ng/mL to an upper limit of about 80-100 to exist for %TS levels up to about 50-55% both diabetes ng/mL as observed in this study. This threshold ferritin level risk [13] and CVD outcomes [21]. coincides with levels associated with increased T2D risk observed epidemiologically [13] as well as adverse outcomes REVISED NORMAL RANGES FOR IRON STATUS observed in clinical trial data [17, 21]. AND LONGEVITY Fig. (2) shows proportionately increasing hemoglobin The revised normal ranges for these biomarkers of iron and %TS levels up to about 20-25%TS. Rising %TS levels metabolism suggested here bear striking resemblance to gradually plateau above an inflection point of about 25% in other estimates of disease risk in relation to increasing fer- relation to corresponding hemoglobin levels. Transferrin ritin and %TS levels. Population mean ferritin levels vary saturation levels below about 20-25% may be consistent with considerably by age, race and sex while %TS levels appear transport of iron as holotransferrin for erythropoiesis [14]. to be relatively stable [20, 21, 27]. Mean ferritin levels up to Levels above about 25%TS imply a possible switch to incor- approximately 60-70 ng/mL in the late teens in males and Iron Measures, Diabetes and Cardiovascular Risk Current Diabetes Reviews, 2017, Vol. 13, No. 0 5 females associate with low disease risk [17, 23-25]. Popula- mean ferritin level for the total cohort at entry was 301 tion levels in males increase at a rate of about 4 to 5 ng/mL ng/mL. Levels remained unchanged during follow-up in con- per year and plateau at a mean of about 140-150 ng/mL as- trol subjects but declined to a mean of 36 ng/mL (p<0.001) sociated with increasing disease risk by middle-age [20]. in subjects randomized to dietary iron restriction. Iron reduc- Cross-sectional measures of ferritin decline gradually with tion was associated with a significant reduction in the pri- increasing age to range at about 80-100 ng/mL suggesting mary study endpoint, p<0.01, and significantly reduced risk that such lower levels select for greater longevity [20, 24, of doubling of the serum , p<0.02. 25] as illustrated in CSP 410 outcomes data [17]. Low risk Ogawa, et al [25] reported clinical outcome data from a mean ferritin levels below about 50 ng/mL in pre- cohort of patients on chronic hemodialysis, 21% of whom menopausal women rise after menopause to plateau at ap- had diabetic nephropathy. Ferritin levels between about 30 proximately 80-100 ng/mL about 30 years earlier in women ng/mL and 80 ng/mL were associated with significantly im- compared to men [20]. The increase in ferritin levels above proved long-term survival (p=0.013) compared to survival this threshold associate with increased post-menopausal dis- with higher or lower ferritin levels and hemoglobin status. ease risk in women [46]. Occurrence of ferritin levels in this range at an earlier age in women correspond to their in- Strong continuous statistical associations exist between creased risk for T2D above this threshold [13]. Higher fer- ferritin levels and insulin resistance (IR) measured typically ritin levels relate to increasingly impaired beta cell function, by the Homeostasis Model Assessment of Insulin Resistance insulin resistance and metabolic abnormalities characterizing (HOMA-IR) assay [51-54]. Correlations between IR and diabetes [5, 47, 48]. ferritin levels above about 99 ng/mL in men and above about 86 ng/mL in women have been reported [50]. Kim et al [54] Ferritin levels below about 80-100 ng/mL relate to mini- showed significant positive statistical associations between mal disease and maximum longevity observed in other stud- levels of insulin, fasting glucose and HOMA-IR with ferritin ies [17, 20, 22-25]. Individuals in the elderly Framingham levels above 98 ng/mL in post-menopausal women (p<0.001 Heart Study cohort having a mean age of 75 (age range 68- for each comparison). Relationships between iron measures 93) but lacking a clinical disease diagnosis had mean ferritin and HOMA-IR were tracked longitudinally in a cohort of levels of 86 ng/mL whereas members of the same elderly otherwise healthy non-diabetic women followed from pre- cohort having a clinical disease diagnosis had a slightly menopause to one-year post-menopause [55]. During meno- higher mean ferritin level of 94 ng/mL [22]. Percent TS lev- pausal transition, fasting glucose (p=0.05) ferritin (p<0.01) els in the Framingham Heart Study cohort were lower in and HOMA-IR (p=0.022) increased significantly. Mean pre- individuals without a clinical diagnosis, 25%TS, compared menopausal ferritin levels were 69.5 ng/mL and mean post- to individuals having a clinical disease diagnosis, 35%TS. menopausal ferritin levels 128.8 ng/mL, resembling low ver- Based on data shown in Figure 3, 25%TS would be consid- sus increased risk levels observed in the present and other ered within the low risk range and 35%TS, while higher, studies [20,22,23]. Improved IR with removal of a single would remain in the compensated range. Data from a cohort unit of blood (500 mL, estimated to reduce ferritin levels by of males from the Mediterranean region of Europe over age about 50 ng/mL in a healthy individual) reinforces the con- 80 having relatively low morbidity and mortality had mean cept that IR risk may be a continuum related to the ferritin ferritin levels of 68 ng/mL [23]. These levels were signifi- level [56]. cantly lower than levels found in a male population of simi- lar age from The Netherlands having significantly greater Relationships between iron measures and T2D risk morbidity associated with mean ferritin levels of 137 ng/mL. [57,58] are relevant to vulnerable populations predisposed Adjusted mean %TS levels were 20% in the healthier Medi- metabolically to diabetes associated with overall increased terranean population versus 34% in the sicker Northern ferritin levels [21,27,58-60]. Living at high altitude corre- European population. Based on our findings, the healthier lates with both reduced diabetes risk [61] along with lower Mediterranean population having mean %TS levels of 20% ferritin levels [62]. Treatment of diabetes with the oral hypo- would be within the low risk range while the higher risk glycemic agent, metformin, significantly reduces risk of dia- Northern Europeans of the same age having mean %TS lev- betic co-morbidities [63-70]. The anti-diabetic properties of els of 35%, while slightly higher, would remain in the com- metformin may be linked to the metal binding properties of pensated range. Residents of the Mediterranean region are this drug [71]. A 3-month course of metformin reduced fer- known to have a lower risk of T2D [49], vascular disease ritin levels by about 50% (p<0.001) and improved insulin and malignancy [23]. These estimates resemble NHANES III resistance in a cohort of women with polycystic ovary syn- data showing a similar pattern for ferritin levels in free- drome [72]. living older age individuals [20]. LIMITATIONS OF THIS STUDY REVISED NORMAL RANGES FOR IRON MEAS- CSP 410 focused on clarification of the role of iron ex- URES AND DIABETES RISK cess in the pathogenesis of CVD that may be potentially The validity of the revised normal values for measures of linked to multiple factors that were not investigated in this iron status reported here is supported by several studies of clinical trial. Free-living CVD patients with and without dia- T2D risk. A prospective randomized interventional trial in betes likely have various co-morbidities or more advanced T2D examined effects of rigorous dietary iron restriction disease associated with higher ferritin levels than are repre- (total n = 191, evaluable n = 170) on development of end sented in this cohort [73]. Elevated ferritin levels may relate stage renal disease requiring transplantation or all-cause to increase in iron stores, inflammatory response, or a com- mortality over a mean follow-up of about 4 years [50]. The bination of both. The plots obtained from the present partici- 6 Current Diabetes Reviews, 2017, Vol. 13, No. 0 Zacharski et al. pants may not be characteristic of the general population. optimal management of diseases of iron excess by the Loess curves reported here were derived from a population American Association for the Study of Liver Disease [92]. consisting almost entirely of males. Curves derived from a population of females may also differ from those obtained in CONCLUSION CSP 410. Measures of hepcidin, NTBI [74-76], , insulin resistance, inflammation and possibly The wide range of putatively “normal” serum ferritin other variables were not available in this study. T2D risk levels, 30-300 ng/mL in men and 15-200 ng/mL in women, a may also be attributable to socio-economic factors, environ- 10-fold variation, represent extreme variations that do not mental contaminants [77-79] or Western Lifestyles [80]. resemble the range of normal values for other laboratory These require further investigation of their effects on iron metabolic measures [2, 24]. Revised normal values as re- biomarkers and associated iron metabolism. flected by relationships between ferritin, %TS and hemoglo- bin observed in the present study also associate with epide- miologic and clinical trial data predicting favorable clinical STRENGTHS OF THIS STUDY outcomes. Optimal (minimum risk) serum ferritin levels ap- Ferritin levels at entry to CSP 410 were a randomization pear to range from a low of about 15 ng/ml to about 80-100 variable considered a “signal” of iron metabolism conveying ng at an upper limit. Data analyzed here for %TS levels re- prognostic information. Ferritin levels at entry and intent-to- semble the range of normal as accepted commonly having a treat ferritin levels during follow-up were interpreted at face lower limit below about 15% and an upper limit of about value to determine the amount of blood to be withdrawn by 55%. We suggest that %TS levels up to about 25% may re- phlebotomy without attempting to differentiate between flect iron transport for erythropoiesis. Ferritin in relation to acute phase and non-acute phase levels. Strengths that render %TS levels ranging between about 25% and 55% may re- data on iron biomarkers suitable for the present analysis in- flect compensation by transferrin scavenging of iron or clude study’s prospective, randomized single-blinded design, NTBI up to ferritin levels of about 100 ng/mL. Ferritin and and longitudinal analysis of a wide range of levels of the %TS levels within these ranges appear to associate with variables of interest. Iron reduction was based on levels of lower T2D [13] and CVD risk [21]. the primary variable of interest (the serum ferritin level) as modified with randomization to calibrated phlebotomy [17, LIST OF ABBREVIATIONS 21]. Ferritin levels in this study are observed commonly in the general population and were subject to experimental CSP 410 = VA Cooperative Studies Program #410 variation [17] as well as natural variation in free-living popu- CVD = cardiovascular disease lations [17, 20]. Diabetes status was also a randomization variable; similar numbers of diabetics and non-diabetics FeAST = The Iron and Atherosclerosis Study, CSP 410 were randomized to iron reduction as compared to control [27-30]. Ferritin and %TS levels reported here were similar GDM = gestational diabetes mellitus between diabetic and non-diabetic entrants while hemoglo- HOMA-IR = Homeostasis Model Assessment of Insu- bin levels were significantly lower in diabetics. Loess meth- lin Resistance odology identified break points in a continuum of ferritin and %TS levels that corresponded to clinical risk [13, 17, HDL = High density lipoprotein cholesterol 21]. Cut points in Loess curves may reflect an apparent tran- hsCRP = high sensitivity C-reactive protein sition from physiologic to pathologic iron biomarker levels that appeared to be similar between participants with or IRB = Institutional review board without T2D. Thus, the demonstrated quantitative relation- LDL = low density lipoprotein cholesterol ships for ferritin and %TS levels that distinguish physiologic from non-physiologic ferritin levels of these biomarkers or IL6 = Interleukin 6 iron metabolism may facilitate clinical correlation of obser- Loess = Locally weighted scatter plot smoothing vations on iron-mediated pathways and risk of T2D [81-83], its complications [84-86] and its co-morbidities [87-90]. Iron NHANES = National Health And Nutrition Examina- biomarker values reported here might be extrapolated to co- tion Survey morbidities of diabetes besides CVD. For example, malig- NTBI = Non-transferrin bound iron nancy may have a similar hypothetical pathophysiologic TNF = Tumor necrosis factor narrative, i.e., damage to cells and tissues caused by iron- catalyzed oxygen free radical generation [24, 91]. CSP 410 TS = transferrin saturation data reported previously showed reduced risk of new visceral T2D = Type 2 diabetes malignancy (p=0.036), and all-cause and cancer-specific mortality (p=0.009) with randomization to iron reduction VA = Veterans Administration [92]. Mean follow-up ferritin levels in entrants not develop- ing cancer as compared to those developing cancer were 76.4 AUTHOR CONTRIBUTIONS ng/mL versus 127.1 ng/mL respectively (p=0.017). These L.R.Z. and R.G.D drafted the manuscript, G.S. and B.K. ferritin levels also correspond to lower versus higher risk C. performed the statistical analyses and prepared the fig- values as suggested by the present report. Additionally, fer- ures. L.R.Z., G.S., B.K.C. and R.G.D. contributed to writing ritin levels below about 100 ng/mL are recommended for and editing the manuscript, and approved the final manu- script for submission. Iron Measures, Diabetes and Cardiovascular Risk Current Diabetes Reviews, 2017, Vol. 13, No. 0 7

ETHICS APPROVAL AND CONSENT TO PARTICI- markedly increased ferritin concentrations: general population PATE study and metaanalysis. Clin Chem 2014;60:1419-1428. [13] Podmore C, Meidtner K, Schulze MB, et al. The Association of Not applicable. Multiple Biomarkers of Iron Metabolism and Type 2 Diabetes: The EPIC-InterAct Study. Diabetes Care 2016;39: doi 10.2337/dc15- 0257. HUMAN AND ANIMAL RIGHTS [14] Elsayed ME, Sharif MU, Stack AG. Transferrin saturation: A body iron biomarker. Adv Clin Chem 2016;75:71-97. doi: No Animals/Humans were used for studies that are base 10.1016/bs.acc.2016.03.002. Epub 2016 May 6. of this research. [15] Jung CH, Lee MJ, Hwang JY, et al. Elevated serum ferritin level is associated with the incident type 2 diabetes in healthy Korean men: a 4 year longitudinal study. PLoS One. 2013; 8: e75250. doi: CONSENT FOR PUBLICATION 10.1371/journal.pone.0075250. eCollection 2013. Not applicable. [16] Aregbesola A, Voutilainen S, Virtanen JK, Mursu J, Tuomainen TP. Body iron stores and the risk of type 2 diabetes in middle-aged men. Eur J Endocrinol 2013;169:247-53. doi: 10.1530/EJE-13- CONFLICT OF INTEREST 0145. [17] Zacharski LR, Shamayeva G, Chow, B.K. Effect of controlled The authors confirm that this article content has no con- reduction of body iron stores on clinical outcomes in peripheral ar- flict of interest. terial disease. Am Heart J 2011;162: 949-957.e1. [18] DePalma RG, Hayes VW, Chow BK, et al. Ferritin levels, inflam- matory biomarkers and mortality in peripheral arterial disease: A ACKNOWLEDGEMENTS sub-study of the Iron (Fe) and Atherosclerosis Study (FeAST) The Cooperative Studies Program (CSP 410) of the De- Trial. J Vasc Surg 2010;51: 1498-1503. [19] DePalma RG, Hayes VW, Cafferata HT, et al. Cytokine signatures partment of Veterans Affairs Office of Research and Devel- in atherosclerotic claudicants. J Surg Res 2003;111: 215-221. opment, Clinical Science Research & Development Service [20] Zacharski LR, Ornstein DL, Woloshin S, Schwartz LM. Associa- supported The Iron and Atherosclerosis Study (FeAST). The tion of age, sex, and race with body iron stores in adults: analysis opinions expressed are those of the authors and not those of of NHANES III data. Am Heart J 2000;140: 98-104. [21] Zacharski LR, Shamayeva G, Chow BK, DePalma RG. Racial the government of the United States or the Department of Differences in Iron Measures and Outcomes Observed during an Veterans Affairs. Iron Reduction Trial in Peripheral Arterial Disease. J Health Care Poor Underserved 2015; 26: 243-59. doi: 10.1353/hpu.2015.0026. REFERENCES [22] Fleming DJ, Jacques PF, Tucker KL, et al. Iron status of the free- living, elderly Framingham Heart Study cohort: an iron-replete [1] Swaminathan S, Fonesca VA, Alam MG, Shah SV. The role of iron population with a high prevalence of elevated iron stores. Am J in in diabetes and its complications. Diabetes Care 2007; 30: 1926- Clin Nutr 2001; 73:638-46. 1933. [23] Buijsse B, Feskens EJ, Moschandreas J, et al. Oxidative stress, and [2] Simcox JA, McClain DA. Iron and Diabetes Risk. Cell Metab iron and antioxidant status in elderly men: differences between the 2013; 17: 329-341. Mediterranean south (Crete) and northern Europe (Zutphen). Eur J [3] Hansen JB, Tonnesen MF, Madsen AN, et al. Divalent metal trans- Cardiovasc Prev Rehabil 2007;14: 495-500. porter 1 regulates iron-mediated ROS and pancreatic beta cell fate [24] Silva B, Faustino P. An overview of molecular basis of iron me- in response to cytokines. Cell Metab 2012;16:449-61. tabolism regulation and the associated pathologies. Biochim Bio- [4] Wlazlo N, van Greevenbroek MM, Ferreira I, et al. Iron metabo- phys Acta 2015; 1852:1347-59. doi: 10.1016/j.bbadis.2015.03.011. lism is prospectively associated with insulin resistance and glucose [25] Ogawa C, Tsuchiya K, Kanda F, Maeda T. Low levels of serum intolerance over a 7-year follow-up period: the CODAM study. ferritin lead to adequate hemoglobin levels and good survival in Acta Diabetol 2015;52: 337-348. hemodialysis patients. Am J Nephrol 2014; 40: 561-70. doi: [5] Fernández-Real JM, McClain D, Manco M. Mechanisms linking 10.1159/000370317. Epub 2015 Jan 10. glucose homeostasis and iron metabolism toward the onset and [26] Cleveland WS, Devlin SJ. Locally weighted regression: an ap- progression of type 2 diabetes. Diabetes Care 2015; 3: 2169-2176. proach to regression analysis by local fitting. J Amer Stat Assn [6] Orban E, Schwab S, Thorand B, Huth C. Association of iron indi- 1988;83: 596–610. ces and type 2 diabetes: a meta-analysis of observational studies. [27] Zacharski LR, Shamayeva G, Chow BK, DePalma RG. Racial Diabetes Metab Res Rev 2014; 30: 372-394. health disparities, and variant red cell and iron homeostasis. [7] Fu S, Li F, Zhou J, Liu, Z. The relationship between Body iron J Health Care Poor Underserved 2016; 27:741-61. status, iron intake and gestational diabetes: A systematic review [28] Zacharski LR, Chow BK, Howes PS, et al. Reduction of Iron and meta-analysis. Medicine (Baltimore) 2016; 95: e2383. Stores and Cardiovascular Outcomes in Patients with Peripheral [8] Zhao Z, Li S, Liu G, et al. Body iron stores and heme-iron intake in Arterial Disease. JAMA 2007; 297: 603-10. relation to risk of type 2 diabetes: a systematic review and meta- [29] Zacharski LR, Chow, BK, Howes PS, Shamayeva G, Lavori PW. analysis. PLoS One 2012; 7: e41641. Implementation of an iron reduction protocol in patients with pe- [9] Bao W, Rong Y, Rong S, Liu, L. Dietary iron intake, body iron ripheral vascular disease: VA Cooperative Study #410: The Iron stores, and the risk of type 2 diabetes: a systematic review and and Atherosclerosis Study. Am Heart J 2004; 148: 386-92. meta-analysis. BMC Med 2012; 10: 119. doi: 10.1186/1741-7015- [30] Zacharski LR, Chow B, Lavori PW, et al. The Iron (Fe) and athero- 10-119. sclerosis study (FeAST). I. A pilot study of reduction of body iron [10] Khambalia AZ, Aimone A, Nagubandi P, et al. High maternal iron stores in atherosclerotic peripheral vascular disease. VA Coopera- status, dietary iron intake and iron supplement use in pregnancy tive Study #410. Am Heart J 2000; 139:337-345. and risk of gestational diabetes mellitus: a prospective study and [31] Ghasemi A, Zahediasl S. Normality Tests for Statistical Analysis: systematic review. Diabet Med 2015; Dec 16. doi: A Guide for Non-Statisticians. Int J Endocrinol Metab 10.1111/dme.13056. 2012;10:486-9. [11] Kunutsor SK, Apekey TA, Walley J, Kain K. Ferritin levels and [32] Wilcoxon F. Individual Comparisons by Ranking Methods. Bio- risk of type 2 diabetes mellitus: an updated systematic review and metrics Bulletin 1945; 1:80-83. meta-analysis of prospective evidence. Diabetes Metab Res Rev [33] Cleophas TJ, Zwinderman AH. The Assessment of Non-Linear 2013; 29: 308-18. Effects in Clinical Research. Pharmacology & Pharmacy 2012; 3: [12] Ellervik C, Marott JL, Tybjærg-Hansen A, Schnohr P, Nordest- 139-147. http://dx.doi.org/10.4236/pp.2012.32021Accessed: Feb- gaard BG. Total and cause-specific mortality by moderately and ruary 14, 2016 [34] Loess methodology: http://www.statisticshowto.com/lowess- smoothing/ Accessed: February 14, 2016 8 Current Diabetes Reviews, 2017, Vol. 13, No. 0 Zacharski et al.

[35] Loess regression curve [56] Facchini FS. Effect of phlebotomy on plasma glucose and insulin fitting:http://www.statsdirect.com/help/content/nonparametric_met concentrations. Diabetes Care 1998; 21: 2190. hods/loess.htm Accessed: February 14, 2016 [57] Huth C, Beuerle S, Zierer A, et al. Biomarkers of iron metabolism [36] Zhu J, Xie J. Nonparametric Variable Selection for Predictive are independently associated with impaired glucose metabolism Models and Subpopulations in Clinical Trials. J Biopharm Stat. and type 2 diabetes: the KORA F4 study. Eur J Endocrinol 2015; 2015; 25: 781–794. doi: 10.1080/10543406.2014.920861 173: 643-53. doi: 10.1530/EJE-15-0631. Epub 2015 Aug 20. [37] Loess regression graphics and clinical trial data display: [58] Kershaw KN, Pender AE. Racial/Ethnic Residential Segregation, https://support.sas.com/documentation/cdl/en/statug/63033/HTML/ Obesity, and Diabetes Mellitus. Curr Diab Rep 2016; 16:108. default/viewer.htm#statug_loess_sect002.htm Accessed: February [59] Baldwin HJ, Green AE, Spellar KM, Arthur PJ, Phillips HG, Patel 14, 2016 JV. Tipping the balance: Haemoglobinopathies and the risk of dia- [38] Warram JH, Scott LJ, Hanna LS, et al. Progression of microalbu- betes. World J Diabetes 2016; 7: 8-13. doi: 10.4239/wjd.v7.i1.8. minuria to proteinuria in type 1 diabetes: nonlinear relationship [60] Zimmermann MB, Fucharoen S, Winichagoon P, et al. Iron me- with . Diabetes 2000; 49: 94-100. tabolism in heterozygotes for hemoglobin E (HbE), alpha- http://dx.doi.org/10.2337/diabetes.49.1.94. thalassemia 1, or beta-thalassemia and in compound heterozygotes [39] Nirantharakumar K, Hemming K, Narendran P, Marshall T, Cole- for HbE/beta-thalassemia. Am J Clin Nutr 2008; 88: 1026-31. man JJ. A Prediction Model for Adverse Outcome in Hospitalized [61] Woolcott OO, Castillo OA, Gutierrez C, Elashoff RM, Stefanovski Patients With Diabetes. Diabetes Care 2013; 36: 3566-3572. D, Bergman RN. Inverse association between diabetes and altitude: http://dx.doi.org/10.2337/dc13-0452. a cross-sectional study in the adult population of the United States. [40] Nissen SE, Murat Tuzcu EM, Libby P, et al, for the CAMELOT Obesity 2014; 22: 2080-90. doi: 10.1002/oby.20800. Epub 2014 Investigators. Effect of Antihypertensive Agents on Cardiovascular May 28. Events in Patients With Coronary Disease and Normal Blood Pres- [62] Cook JD, Boy E, Flowers C, Daroca Mdel C. The influence of sure. The CAMELOT Study: A Randomized Controlled Trial. high-altitude living on body iron. Blood 2005; 106: 1441-6. Epub JAMA 2004; 292: 2217-2225. doi:10.1001/jama.292.18.2217. 2005 May 3. [41] Kulloa IJ, Baileyc KR, Kardiad SLR, Mosleye TH Jr, Boerwinklef [63] Fung CS, Wan EY, Wong CK, Jiao F, Chan AK. Effect of met- E, Turnera ST. Ethnic differences in peripheral arterial disease in formin monotherapy on cardiovascular diseases and mortality: a the NHLBI Genetic Epidemiology Network of Arteriopathy retrospective cohort study on Chinese type 2 diabetes mellitus pa- (GENOA) study. Vasc Med 2003; 8: 237–242. tients. Cardiovasc Diabetol 2015; 14: 137. doi: 10.1186/s12933- [42] Angelousi A, Larger E. Anaemia, a common but often unrecog- 015-0304-2. nized risk in diabetic patients: a review. Diabetes Metab 2015; [64] Higurashi T, Hosono K, Takahashi H, et al. Metformin for chemo- 41:18-27. doi: 10.1016/j.diabet.2014.06.001. Epub 2014 Jul 17. prevention of metachronous colorectal adenoma or polyps in post- [43] Lee DH, Zacharski LR, Jacobs DR, Jr. Comparison of the serum polypectomy patients without diabetes: a multicentre double-blind, ferritin and percentage of transferrin saturation as exposure markers placebo-controlled, randomised phase 3 trial. Lancet Oncol 2016; of iron-driven oxidative stress-related disease outcomes. Am Heart 17: 475-83. doi: 10.1016/S1470-2045(15)00565-3. Epub 2016 Mar J 2006;151: 1247.e1-7. 3. [44] Ponka P, Beaumont C, Richardson DR. Function and regulation of [65] Chan AT. Metformin for cancer prevention: a reason for optimism. transferrin and ferritin. Semin Hematol 1998; 35:35-54. Lancet Oncol 2016; 17: 407-9. doi: 10.1016/S1470- [45] Adams PC, Reboussin DM, Press RD, et al. Biological variability 2045(16)00006-1. Epub 2016 Mar 3. of transferrin saturation and unsaturated iron binding capacity. Am [66] Wei M, Liu Y, Wang C, et al. The need for differentiating diabetes- J Med 2007; 120: 999e1–999e7. specific mortality from total mortality when comparing metformin [46] Brand JS, van der Schouw YT, Onland-Moret NC, et al. Age at with insulin regarding cancer survival. Acta Diabetol 2016 Oct 28. menopause, reproductive life span, and type 2 diabetes risk: results [Epub ahead of print]. from the EPIC-InterAct study. Diabetes Care 2013; 36: 1012-9. [67] Stopsack KH, Ziehr DR, Rider JR, Giovannucci EL. Metformin doi: 10.2337/dc12-1020. Epub 2012 Dec 10. and prostate cancer mortality: a meta-analysis. Cancer Causes Con- [47] Bonfils L, Ellervik C, Friedrich N, et al. Fasting serum levels of trol. 2016; 27:105-13. doi: 10.1007/s10552-015-0687-0. ferritin are associated with impaired pancreatic beta cell function [68] Chan KM, Kuo CF, Hsu JT, et al. Metformin confers risk reduction and decreased insulin sensitivity: a population-based study. Diabe- for developing hepatocellular carcinoma recurrence after liver re- tologia 2015;58: 523-533. section. Liver Int 2016; Oct 24. doi: 10.1111/liv.13280. [Epub [48] Hansen JB, Moen IW, Mandrup-Poulsen T. Iron: the hard player in ahead of print]. diabetes pathophysiology. Acta Physiol (Oxf) 2014;210:717-32. [69] Sun GE, Wells BJ, Yip K, et al. Gender-specific effects of oral [49] de Lorgeril M, Salen P. The Mediterranean diet to prevent type 2 hypoglycaemic agents on cancer risk in type 2 diabetes mellitus. diabetes and its complications. Chapter 31 in: Preedy VR, Watson Diabetes Obes Metab 2014; 16: 276-83. doi: 10.1111/dom.12231. RR, The Mediterranean Diet: An Evidence-Based Approach. El- Epub 2013 Dec 5. sevier, Amsterdam, 2015; pp 337-342. [70] Xu H, Chen K, Jia X, et al. Metformin Use Is Associated With [50] Facchini FS1, Saylor KL. A low-iron-available, polyphenol- Better Survival of Breast Cancer Patients With Diabetes: A Meta- enriched, carbohydrate-restricted diet to slow progression of dia- Analysis. Oncologist 2015; 20: 1236-44. doi: betic nephropathy. Diabetes 2003; 52:1204-9. 10.1634/theoncologist.2015-0096. Epub 2015 Oct 7. [51] Pham NM, Nanri A, Yi S, et al. Serum ferritin is associated with [71] Logie L, Harthill J, Patel K, et al. Cellular responses to the metal- markers of insulin resistance in Japanese men but not in women. binding properties of metformin. Diabetes 2012; 61: 1423-33. doi: Metabolism 2013; 62: 561-7. doi: 10.1016/j.metabol.2012.07.025. 10.2337/db11-0961. Epub 2012 Apr 9. Epub 2012 Oct 27. [72] Behradmanesh S, Ranjbar Omrani GH, Ghazanfarpour F, Barada- [52] Kim CH, Kim HK, Bae SJ, Park JY, Lee KU. Association of ele- ran A. Effect of metformin on serum ferritin level in women with vated serum ferritin concentration with insulin resistance and im- polycystic ovary syndrome. paired glucose metabolism in Korean men and women. Metabolism [73] Red Crescent Med J 2011;13:487-92. Epub 2011 Jul 1. 2011; 60: 414–420. [74] Jenq CC, Hsu CW, Huang WH, Chen KH, Lin JL, Lin-Tan DT. [53] Alnahal AA, Tahan M, Fathy A, Fathy T. Effect of deferoxamine Serum ferritin levels predict all-cause and infection-cause 1-year therapy on insulin resistance in end-stage renal disease patients mortality in diabetic patients on maintenance hemodialysis. Am J with iron overload. Saudi J Kidney Dis Transpl 2014; 25: 808-13. Med Sci 2009; 337:188-94. doi: 10.1097/MAJ.0b013e31818d8bbe. [54] Kim MK, Chon SJ, Jung YS, et al. The Relationship between Se- [75] de Swart L, Hendriks JC, van der Vorm LN, et al. Second interna- rum Ferritin Levels and Insulin Resistance in Pre- and Postmeno- tional round robin for the quantification of serum non-transferrin- pausal Korean Women: KNHANES 2007-2010. PLoS One 2016; bound iron and labile plasma iron in patients with iron-overload 11: e0157934. doi: 10.1371/journal.pone.0157934. eCollection disorders. Haematologica 2016;101: 38-45. 2016. Erratum in: PLoS One. 2016;11(7):e0160300. [76] Lee DH, Liu DY, Jacobs DR Jr, et al. Common presence of non- [55] Kim C, Nan B, Kong S, Harlow S. Changes in iron measures over transferrin-bound iron among patients with type 2 diabetes. Diabe- menopause and associations with insulin resistance. J Women’s tes Care 2006; 29: 1090-1095. Health 2012; 21: 872-7. doi: 10.1089/jwh.2012.3549. Epub 2012 [77] Aljwaid H, White DL, Collard KJ, Moody AJ, Pinkney JH. Non- Jun 25. transferrin-bound iron is associated with biomarkers of oxidative Iron Measures, Diabetes and Cardiovascular Risk Current Diabetes Reviews, 2017, Vol. 13, No. 0 9

stress, inflammation and endothelial dysfunction in type 2 diabetes. [87] Gaudio A, Privitera F, Pulvirenti I, Canzonieri E, Rapisarda R, J Diabetes Complications. 2015;29:943-9. doi: Fiore CE. The relationship between inhibitors of the Wnt signalling 10.1016/j.jdiacomp.2015.05.017. Epub 2015 Jun 3. pathway (sclerostin and Dickkopf-1) and carotid intima-media [78] La Merrill MA. The economic legacy of endocrine-disrupting thickness in postmenopausal women with type 2 diabetes mellitus. chemicals. Lancet Diabetes Endocrinol 2016; Diab Vasc Dis Res 2014; 11: 48-52. doi: doi.org/10.1016/52213-8587(16) 30279-0. 10.1177/1479164113510923. Epub 2013 Nov 12. [79] Franks PW, McCarthy MI. Exposing the exposures responsible for [88] Ma Y, Hébert JR, Balasubramanian R, et al. All-cause, cardiovas- type 2 diabetes and obesity. Science 2016; 354: 69-73. cular, and cancer mortality rates in postmenopausal white, black, [80] Trasande L, Lampa E, Lind L, Lind PM. Population attributable Hispanic, and Asian women with and without diabetes in the risks and costs of diabetogenic chemical exposures in the elderly. J United States: the Women's Health Initiative, 1993-2009. Am J Epidemiol Community Health doi:10.1136/jech-2016-208006. Epidemiol 2013;178: 1533-41. doi: 10.1093/aje/kwt177. Epub [81] Ghio AJ, Schreinemachers DM. Heme Oxygenase Activity Corre- 2013 Sep 17. lates with Serum Indices of Iron Homeostasis in Healthy Non- [89] Stroh M, Swerdlow RH, Zhu H. Common defects of mitochondria smokers. Biomark Insights 2016; 11:49-54. doi: and iron in neurodegeneration and diabetes (MIND): a paradigm 10.4137/BMI.S36226. eCollection 2016. worth exploring. Biochem Pharmacol 2014; 88: 573-83. doi: [82] Andrews M, Leiva E, Arredondo-Olguín M. Short repeats in the 10.1016/j.bcp.2013.11.022. Epub 2013 Dec 19. heme oxygenase 1 gene promoter is associated with increased lev- [90] Cheung CL, Cheung TT, Lam KS, Cheung BM. High ferritin and els of inflammation, ferritin and higher risk of type-2 diabetes mel- low transferrin saturation are associated with pre-diabetes among a litus. J Trace Elem Med Biol 2016; 37: 25-30. doi: national representative sample of U.S. adults. Clin Nutr 2013; 32: 10.1016/j.jtemb.2016.06.001. Epub 2016 Jun 2. 1055-60. doi: 10.1016/j.clnu.2012.11.024. Epub 2012 Dec 28. [83] Tiwari S, Ndisang JF. The heme oxygenase system and type-1 [91] Weinberg ED. The role of iron in cancer. Eur J Cancer Prev 1996; diabetes. Curr Pharm Des 2014; 20:1328-37. 5: 19-36. [84] Abraham NG, Junge JM, Drummond GS. Translational Signifi- [92] Zacharski LR, Chow BK, Howes PS, et al. Decreased cancer risk cance of Heme Oxygenase in Obesity and Metabolic Syndrome. after iron reduction in patients with peripheral arterial disease: re- Trends Pharmacol Sci 2016; 37: 17-36. doi: sults from a randomized trial. J Natl Cancer Inst 2008; 100: 996- 10.1016/j.tips.2015.09.003. Epub 2015 Oct 26. 1002. doi: 10.1093/jnci/djn209. Epub 2008 Jul 8. [85] Pagnin E, Maiolino G, Calò LA. Heme oxygenase-1 in type 2 dia- [93] Bacon BR, Adams PC, Kowdley KV, Powell LW, Tavill AS. betes: from cell first-line defense to early marker of diabetic neph- American Association for the Study of Liver Diseases. Diagnosis ropathy. Minerva Med 2016; 107: 123-4. and management of hemochromatosis: 2011 practice guideline by [86] Negi G, Nakkina V, Kamble P, Sharma SS. Heme oxygenase-1, a the American Association for the Study of Liver Diseases. Hepa- novel target for the treatment of diabetic complications: focus on tology 2011; 54: 328-43. diabetic peripheral neuropathy. Pharmacol Res 2015; 102: 158-67. doi: 10.1016/j.phrs.2015.09.014. Epub 2015 Oct 22.